AI Should Support Clinicians, Not Replace Them: Vivek Subbiah, MD

Artificial intelligence (AI) should assist clinicians in clinical decision-making, not replace them, says Vivek Subbiah, MD, chief of early-phase drug development at Sarah Cannon Research Institute, in the third and final part of an interview at the Patient-Centered Oncology Care^® conference in Nashville, Tennessee.

He adds that AI has the potential to identify patients most likely to benefit from targeted therapies, helping ensure the right treatment reaches the right patient at the right time.

Watch parts 1 and 2 to learn more about how AI is improving patient care in precision oncology and expanding access to clinical trials.

This transcript has been lightly edited; captions were auto-generated.

Transcript

What policy considerations need attention as AI becomes more prevalent in precision oncology, both in diagnostics and trial matching?

I think we need to vet AI because the challenge is, given its black box warning and all that, we do not know what data it is fed. Again, with large language models, LLMs, we need to make sure that the system is vetted. AI should assist clinicians in making decisions. AI should not be making clinical decisions as of yet, not clinical decisions, not for real patient care; AI should assist us in making decisions.

The current challenge that we have is that the large language models have information not just from peer-reviewed literature, but from science fiction books and all that. Again, it can hallucinate a lot of facts, and it's hard to sift through facts and fiction. Still, you need a clinical oncologist, a precision medicine scientist, to vet all this information. I think we need to get to a time where we have a system that helps us, and the system is vetted thoroughly.

How do you see AI shaping precision oncology over the next 5 years, particularly in making care more equitable?

My hope is that it can help us realize the central mantra of precision oncology, which is delivering the right drug to the right patient at the right time. As I said, we have unprecedented advances in genomically targeted therapies, agents that arm the immune system, and antibody-drug conjugates, but the key to delivering these drugs is to identify those patients who will respond to these therapies.

I think AI can do that at scale, and that's our hope for the next 2, 3, 4 years: that AI can help us identify these patients so that they can benefit from all the advances in precision oncology.